You are on page 1of 29

INTERNATIONAL JOURNAL FOR NUMERICAL AND ANALYTICAL METHODS IN GEOMECHANICS

Int. J. Numer. Anal. Meth. Geomech., 2005; 29:443471


Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/nag.421

Numerical modelling of compensation grouting


above shallow tunnels
C. Wisser1,n,y, C. E. Augarde2 and H. J. Burd3
1

Mott MacDonald, Formerly Department of Engineering Science, Oxford University, U.K.


2
School of Engineering, University of Durham, South Road, U.K.
3
Department of Engineering Science, Oxford University, U.K.

SUMMARY
This paper describes the development of a numerical model for compensation grouting which is a useful
technique for the protection of surface structures from the potentially damaging movements arising from
tunnel construction. Pipes are inserted into the ground between the tunnel and the overlaying structure
from an access shaft. Buildings on the surface are instrumented and movements are carefully monitored.
Once the deformations exceed a certain Trigger Level, grout is injected into the ground to prevent damage.
In the nite element model described here, compensation grouting is modelled by applying an internal
pressure to zero-thickness interface elements embedded in the mesh. An observational algorithm is used,
where the deformations of the surface are monitored and used to control the injection process. Example
analyses of compensation grouting are given for three-dimensional tunnel construction underneath a
greeneld site. Dierent strategies are used to control the injection process and their eectiveness in
preventing surface movement is assessed. The numerical model is shown to replicate general behaviour
expected in the eld and is capable of modelling the control of ground surface movements at a greeneld
site. Copyright # 2005 John Wiley & Sons, Ltd.
KEY WORDS:

nite elements; compensation grouting; tunnels

1. INTRODUCTION
Compensation grouting (CG) is a technique for controlling ground movements caused by the
construction of shallow tunnels. A liquid grout is pumped into the ground through narrow tubes
(Tubes a! Manchette or TAMs) between the advancing tunnel and the ground surface.
Movements occur at the tunnel excavation due to various mechanisms of ground loss. The
intention of the grouting is to prevent any signicant movements from propagating to the
surface, thus reducing the likelihood of damage to overlying structures. Successful CG requires
detailed monitoring of surface movements in conjunction with careful control of grout injection.
It is also a relatively new procedure; the rst published use of CG to control tunnelling
n

Correspondence to: C. Wisser, Mott MacDonald, St. Anne House, Wellesley Road, Croydon CR9 2UL, U.K.
E-mail: claus.wisser@email.de

Contract/grant sponsor: EPSRC

Copyright # 2005 John Wiley & Sons, Ltd.

Received 9 February 2004


Revised 15 November 2004

444

C. WISSER, C. E. AUGARDE AND H. J. BURD

settlements in the U.K. appears to have been in 1994 at Waterloo Station, London [1]. The
technique has since come to wider prominence, following its extensive use during the
construction of the Jubilee Line Extension in London where tunnels were driven through
London Clay close to many prestigious buildings. The detailed mechanics of CG are not well
understood. Researchers are now, however, using physical [2] and analytical modelling [311] to
develop an improved understanding of the mechanics of CG. An improved understanding has
potential economic benets; for example, a realistic model could provide predictions of the
eects of a particular scheme of grouting (i.e. the layout of TAMs) and thus suggest alternative
and cheaper arrangements.
This paper describes the development and use of a three-dimensional nite element model of
fracture grouting for tunnelling in overconsolidated clay. In addition to a number of new nite
element procedures, the model also incorporates a range of strategies for controlling the
modelling of grout injection.

2. PREVIOUS NUMERICAL MODELS OF FRACTURE GROUTING


Fracture grouting is a form of compensation grouting in which a low viscosity grout is injected
into the ground under high pressure to form fracture planes along which the grout ows.
Grouting is usually carried out in three stages. Firstly the ground is conditioned by injection of
grout until a small movement is observed at the ground surface. This prepares the ground for
the second stage of grouting and ensures that any subsequent injections are immediately
eective. Secondly, grout is injected as the tunnel is advanced (a process referred to as
concurrent grouting). This is controlled by close monitoring of movements of the ground
surface or structures on the surface. A third stage of grouting may be needed some time after
completion of the tunnel to compensate for consolidation settlements.
Most published numerical modelling of compensation grouting is restricted to the second
(concurrent) stage where better understanding of the mechanics could lead to the greatest
economic benet. Previous research in this area has generally been based either on the use of a
prescribed strain approach to model grout injection or, alternatively, a prescribed pressure
approach. These two approaches are outlined below.
In a prescribed strain approach, grout injection is simulated by imposing appropriate values
of strain on the elements representing the grouted soil. This type of approach is described by
Nicolini and Nova [12] who adopt a set of anisotropic inelastic strains to represent the grout
injection process. This allows the magnitude of the imposed strain to be dierent in dierent
directions and also to vary spatially. Also, Nicholson et al. [3] describe a model for fracture
grouting in which a specied volumetric strain is applied to elements at a pre-dened grout level.
Similar procedures are described by Mayer and Kudella [4], Falk and Schweiger [7] and
Schweiger and Falk [8].
In a prescribed pressure approach, grout injection is generally modelled by a two-stage
procedure. Firstly, an appropriate numerical scheme is used to represent the eect of the
pressure of the liquid grout on the neighbouring soil. This procedure may require the use of
special grout elements. Secondly, the properties of the grout are changed, once injection is
complete, to simulate the increase in stiness of the grout as it sets. Two-dimensional numerical
schemes adopting this type of approach, in which line elements are used to model layers of grout
formed by fracture grouting, are described in References [5, 6, 10]. A three-dimensional model
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

COMPENSATION GROUTING ABOVE SHALLOW TUNNELS

445

employing this approach is described by Soga et al. [2] and Buchet and Van Cotthem [9]. In
these 3D models, based on the nite dierence program FLAC3D, the injection process is
modelled by the application of internal pressure to the relevant solid elements. After injection is
complete, the stiness of these elements is increased to an appropriate value.
These approaches oer two fundamentally dierent means of modelling grout injection
processes. In both cases, it is necessary to specify, at the start of the analysis, certain details
relating to the way in which grout ows into the ground. In a prescribed strain approach, the
extent of the grouting and the magnitude and spatial variation of the imposed strains must be
dened in advance. In the prescribed pressure approach, however, it may be possible to devise
more detailed models to represent the way in which the grout actually ows in the ground; in
this case fewer assumptions need to be made in advance by the analyst. For example, in heavily
overconsolidated clay (such as London Clay), injected grout tends to form thin horizontal
layers. In cases where the vertical stresses vary with position (e.g. beneath a building) grout
would be expected to ow preferentially towards areas of low vertical stress. A prescribed
pressure approach would allow for the possibility of a numerical procedure in which this
preferential ow of grout is modelled [11]. This removes the need for the analyst to specify full
details of the nal size and shape of the grouted zone before the analysis starts.
In practice, compensation grouting is generally undertaken by pumping a specied volume of
grout into the ground. In a numerical model based on a prescribed strain approach, this volume
is imposed directly on the soil by the specication of an appropriate distribution of strain.
Although in a numerical model based on a prescribed pressure approach the grout injection is
controlled by the application of pressure (and not volume), the possibility does exist of imposing
a specied volume of grout by simply terminating the grout injection process when the volume
of grout injected into the mesh (as indicated by the increase in volume of the grout elements)
reaches the required value.
Prescribed strain approaches, in general, have the considerable advantage of simplicity in
comparison with prescribed pressure approaches. However, prescribed pressure approaches do
appear to have the advantage that, in principle at least, they are capable of more detailed
modelling of the ow of grout into the ground. In spite of the additional complexity involved,
therefore, a prescribed pressure approach is adopted in the analyses described in this paper.

3. PROPOSED MODEL OF GROUT INJECTION


3.1. Background
The grout injection model described in this paper is based on the 3D tunnelling model described
in Reference [13]. The features of this earlier model that are relevant to the current study, are
outlined below.
(a) The soil is assumed to behave in an undrained manner. It is modelled using a multisurface elasto-plastic model [14] which captures the small strain non-linearity that is
typical of overconsolidated clays such as London Clay. It consists of an arbitrary
number of nested von Mises yield surfaces which form cylinders parallel to the space
diagonal in total principal stress space. The inner surfaces move in stress space according
to a set of linear strain hardening rules. When a stress point reaches a yield surface the
stiness is reduced (Figure 1) and those yield surfaces in contact with it follow its stress
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

446

C. WISSER, C. E. AUGARDE AND H. J. BURD

path (Figure 2). It models the non-linearity of clay at small strains, the eect of the recent
stress history on the current behaviour and the memory loss of the recent stress history,
when all surfaces have become fully re-orientated.
Each yield surface is described by two parameters specifying the size of the surface and
the magnitude of the stiness reduction as the yield surface is activated. The size of the
outermost surface is xed and determines the undrained shear strength of the material.
In the calculations described in this paper, the parameters were selected to give a
variation of stiness with strain that is representative of the behaviour of London Clay.
The resulting variation of the tangent stiness against shear strain based on the
parameters in Table I is shown in Figure 3. The undrained strength prole is taken from
data for London Clay given by St. John et al. [15]. The undrained shear strength
increases linearly with depth: su 60 6z kPa; where z is the distance (in m) below the
ground surface. Kim [16] and Ng et al. [17] report that the ratio of small strain shear

Figure 1. Variation of shear stress with shear strain for nested yield surface model [14].

outer surface
(fixed in stress space)

inner yield
surfaces

2
initial state

2
translated surfaces

Figure 2. Nested yield surface model [14].


Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

447

COMPENSATION GROUTING ABOVE SHALLOW TUNNELS

Table I. Parameters used for nested yield surface model (for a description of the soil model and its
parameters see Reference [14]).
Gs0
4

su0

0.49

60 kPa

6 kPa=m

20 kN=m3

3:0  10 kPa=m

3:0  10 kPa

c0a

g0a

Surface

c0a

g0a

0.02
0.04
0.06
0.1
0.15

0.9
0.75
0.5
0.3
0.2

6
7
8
9
}

0.2
0.3
0.5
0.7
}

0.3
0.15
0.05
0.025
}

Surface
1
2
3
4
5

1.2
1.0

G/G0

0.8
0.6
0.4
0.2
0.0
1.00E-04

1.00E-03

1.00E-02

1.00E-01

1.00E+00

1.00E+01

shear strain (%)

Figure 3. Variation of tangent shear modulus with strain (triaxial).

stiness to undrained shear strength G0 =su for London Clay is about 100250. However,
Liu [18] reports numerical diculties when adopting a value of 250. A value of 500 was
therefore used for the analyses in this paper.
(b) The soil is modelled using an unstructured mesh of 10-noded tetrahedral elements. The
analyses were conducted using the nite element program OXFEM [13] which has been
developed at Oxford University, U.K. for the study of problems in geotechnical
engineering. This program deals with non-linear problems using a modied Euler
approach.
It should be noted that in the analyses described by Burd et al. [13] the tunnel was unlined. In
the model described in this paper, however, the tunnel was lined with a thin layer of tetrahedral
continuum elements [11, 19].
The procedure to model tunnel installation is as follows:
1. The nite element mesh is generated to include the volume of soil that is removed during
the tunnelling operation and also a thin tube of continuum elements in the location where
the lining is installed.
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

448

C. WISSER, C. E. AUGARDE AND H. J. BURD

2. Tunnel installation is modelled in incremental lengths. At the start of an installation


increment the properties of the continuum elements representing the appropriate length of
tunnel lining are changed from their initial values (corresponding to soil) to values
appropriate for the modelling of the tunnel lining material.
3. Soil elements within the tunnel lining are removed (using the procedure described by
Brown and Booker [20]).
4. A uniform hoop strain is applied to the tunnel liner. This provides a specied reduction in
cross-sectional area of the tunnel and is the means adopted in the analysis to prescribe the
ground loss. Since the imposed hoop strain is uniform this procedure does not generate
bending moments in the liner (although bending moments may be generated in response to
the stresses applied by the surrounding soil)
The liner consists of a contiguous ring. No attempt is made to model the detailed structure of a
segmental lining.
It should be noted that the calculations described in this paper are focussed on the application
of compensation grouting to greeneld site conditions. Although preliminary studies have also
been carried out on compensation grouting beneath a surface structure (see Reference [11]) these
studies have served principally to demonstrate that, for the case where a building is present, the
problem of controlling the grout injection process becomes substantially more dicult than is
the case for the greeneld site. This paper is therefore devoted to the study of compensation
grouting processes for the case where a building is absent. However, an assessment of the
preliminary studies conducted by Wisser [11] for the case where a building is present is given in
Section 4.4 of this paper.
The model of compensation grouting described in this paper is concerned entirely with the
immediate (undrained) response of the soil to the injection of grout. It is likely that the injection
of grout will lead to the generation of excess pore-pressures in the neighbouring soil and that the
dissipation of these pore pressures will lead to further soil deformations. The study of these
long-term movements, however, is beyond the scope of the current paper.
3.2. Modelling grout injection
The model of CG described in this paper consists of two separate processes: (a) procedures to
model the injection of grout and (b) procedures to control the grout injection process in
response to movements at the ground surface. This section describes the grout injection model.
Procedures to control the grout injection process are described in Section 3.3.
This paper is concerned entirely with fracture grouting in a heavily overconsolidated clay
(e.g. London Clay). In heavily overconsolidated soils, there is evidence both from site
investigations [21] and laboratory tests [22] that, as a consequence of the horizontal in situ stress
being larger than the vertical in situ stress, the grout opens horizontal fracture planes in the soil.
Grout ows along these fracture planes to form horizontal layers. To model this behaviour,
grout elements are generated in the nite element mesh in horizontal planes in the locations
where grout is assumed to ow. The grout injection procedure is modelled by a prescribed
pressure approach as described below. One drawback of the approach used here is that the
extent of the grouting patch has to be dened by the user. A rened model, in which the
interaction between the grout pressure and the in situ stresses in the ground is used to determine
the distance travelled by the grout, is given by Wisser [11]. Further discussion of this advanced
model, however, is beyond the scope of the current paper.
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

COMPENSATION GROUTING ABOVE SHALLOW TUNNELS

449

The grout elements used in the model are based on conventional 12-noded interface elements
as described in Ngo-Tran [23], and as shown in Figure 4. The triangular faces of the grout
elements match the faces of the tetrahedra either side of the grout layer. Values of stiness are
specied in the normal and the two orthogonal shear directions.
To limit the spatial extent of the grout element panels, the edges of the interface elements
coincident with the edge of the panel must be tied together. This could be achieved by altering
the formulation of these particular edge interface elements to include constraints on element
node pairs along these edges. An alternative approach is used here, where the element edges are
joined together by separate tie elements that implement the constraints directly by coupling the
degrees of freedom of the nodes on the edge. This approach avoids the need for dierent types of
grout elements, and simplies mesh generation. Tie elements to join 2D and 3D elements
together have been developed by Houlsby et al. [24] for modelling 2D masonry facades in
tunnelling settlement damage analyses [13]. In the model described in this paper the tie element
formulation was adapted to tie two 3D nodes together to create a patch edge. Further details of
the formulation of these tie elements are given in Reference [11].
The procedure for modelling grout injection, in outline, is as follows:
(a) The initial assignment of high shear and normal stiness to the grout elements to tie
together the gap into which grout will ow. The stiness is selected to ensure that any
additional exibility introduced by the grout elements is minimal.
(b) Modelling of grout injection. This is achieved by reduction of grout element stiness to
zero and application of an appropriate internal pressure to the grout elements.
(c) Control of the grout pressure by the monitoring of surface movements (see Section 3.3).
(d) Grout cure: an appropriately high stiness is assigned to the grout elements. This is
achieved by changing the material parameters in a single calculation step.
To achieve Step (b), an incremental grout pressure Dp is applied repeatedly to increase the
internal grout pressure pg : Details of this procedure are set out below.
Vertical compressive stresses generally exist in the grout elements as a consequence of the
vertical stresses in the neighbouring soil elements (Figure 5(a)). At the start of the grout
injection analysis, the value of the smallest vertical compressive stress sz min at any Gauss point
within the grout panel is determined. Clearly, the applied grout pressure must exceed sz min for
grout to ow into the panel and so the grout pressure, pg0 ; at the start of the grout injection step
is set to pg0 sz min :
9
12

3
11

10

z
x

Figure 4. 12 noded interface element.


Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

450

C. WISSER, C. E. AUGARDE AND H. J. BURD

current extent of
hydrofracture zone
pg1
pi
(z ) min

(z ) min
extent of grout panel
Gauss point i

extent of grout panel

(a)

location

(b)

location

current extent of
hydrofracture zone

z
pg2

p
pg1

extent of grout panel

(c)

location

Figure 5. Calculation of grout pressure across a grout panel: (a) initial stresses (compression positive);
(b) rst grouting increment; and (c) second grouting increment.

An increment of grout pressure Dp is then applied; the applied grout pressure, pg1 ; at the end
of this rst increment is therefore pg1 pg0 Dp (see Figure 5(b)). However, the value of
incremental grout pressure applied at each individual Gauss point will depend on the value of
the pre-existing vertical stress, as shown in Figure 5(b). At the ith Gauss point, the applied
incremental grout pressure Dpi ; is
Dpi pg1  szi

for pg1  szi > 0

Dpi 0

for pg1  szi  0

where szi is the vertical stress at the ith Gauss point at the start of the grout increment.
The values of incremental grout pressure are applied in the analysis by the application of
incremental nodal loads, f ; evaluated from
Z
f
BT Dpi dA
2
A

where B is the appropriate matrix relating nodal displacements to relative displacement for the
grout element and dA is an element of area. The deformations within the mesh, and the resulting
Gauss point stresses, are computed using conventional nite element techniques. Further
increments Dpi are applied using the above procedure until certain measures of movement at the
ground surface fall to acceptable limits.
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

COMPENSATION GROUTING ABOVE SHALLOW TUNNELS

451

The procedure for modelling the injection of grout in the ground, described above, does not
involve the specication of the volume of grout actually injected into the ground. This is a
consequence of the pressure-controlled approach that is adopted. However, in a practical
analysis, it may be desirable to compute the volume of grout that is required to achieve the
required amount of settlement control. This calculation could be achieved relatively easily by
integrating, over the area of each grout patch, the normal displacements developed in the grout
elements. However, this particular calculation was not carried out in the analyses described in
this paper.
3.3. Grouting Strategies
To control the compensation grouting process the displacement of a set of points on the ground
surface (referred to in this paper as Observation Points) are monitored. Appropriate values of
grout pressure are then applied (using procedures described in Section 3.2) to control the
movements of these points. For modelling purposes, a set of rules are required to allow
appropriate values of grout pressures to be computed and applied. The set of rules governing the
acceptable limits of movement are referred to in this paper as a Grouting Strategy.
Implementation of an appropriate Grouting Strategy can lead to considerable additional
complexity in the model for the following reasons. Firstly, to ensure that the requirements of the
Grouting Strategy are satised, a feedback system needs to be implemented to ensure that the
grout pressure is set to an appropriate value. A second source of complexity arises from the need
to formulate each strategy in a way that leads to a stable and unique numerical algorithm. In
practice, decisions on grout pressures and injection locations are generally made by a panel of
experts on the basis of recent data on ground movements. To approach a realistic model, the
combined experience of a panel of experts needs to be encapsulated within an appropriate
numerical procedure. This is a far from straightforward procedure.
A Grouting Strategy is a set of rules that compares the parameters derived from movements of
the Observation Points (referred to as Settlement Control Parameters (SCPs)) with predened
limits known as Trigger Levels and Target Levels. Possible Settlement Control Parameters are,
for example, absolute settlement or deection ratio. A Trigger Level, TR, is the critical value of
SCP at which grouting should begin. It represents the limit at which the magnitude of the ground
movement is about to become unacceptable. The Target Level, TA, is the parameter value at
which grouting can stop, ground surface movements having been reduced to an acceptable limit.
Each increment of grout injection is applied as a single load step in a non-linear analysis (see
Section 3.2), at the end of which deformations are determined and new values of Settlement
Control Parameters derived. If the Target Level is overshot, this analysis step is automatically
discarded, and a smaller grout pressure increment is applied. The new pressure increment is
determined by linear interpolation between the parameter values at the start and end of the
discarded step. A ow chart for the settlement control algorithm is given in Figure 6. Since it is
dicult to reduce the pressure to the level so that the Settlement Control Parameter exactly
meets the specied Target Level, the user also species a tolerance e: Grouting is stopped when
the observed SCP satises: TA  e4SCP4TA e:
The purpose of compensation grouting usually is to control surface deformations to avoid
damage to buildings. Building damage is commonly classied into six categories, from negligible
to very severe according to visual assessment [25] or by simple elastic analysis assuming the
structure to be a deep beam [26]. These categories of damage can be linked to direct
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

452

C. WISSER, C. E. AUGARDE AND H. J. BURD

start

yes

grouting
necessary?

no

apply grouting
pressure
calculate
displacements

yes

target level
met?
yes

no

target level
exceeded?

no

calculate new
grouting pressure

end
Figure 6. Flow chart for grouting calculation.

measurements of building movements to allow assessment to take place during tunnelling


operation. In many cases, however, these calculations are completed before the start of the
project to estimate values of vertical displacement at certain points in the structure that would
lead to excessive damage. These calculated displacement values may then be treated as
settlement limits that should not be exceeded during construction. One example of this approach
occurred in the Jubilee Line project where a limit of 25 mm was applied to all rail structures [27].
Alternatively, it may be of greater importance to check levels of dierential settlement (i.e.
relative movement measured between two points). A common dimensionless measurement of
dierential settlement is deection ratio DR,
D
3
DR
L
where D is the maximum vertical deection measured between two points a distance L apart (see
Figure 7). However, it seems that in practice, reported practical procedures in which
compensation grouting is used to limit deection ratio are less common than those to limit
discrete values of settlement.
Three Grouting Strategies implemented in the numerical model are described below:
Strategy A. Grouting Strategy A uses absolute settlement of the Observation Points as the
Settlement Control Parameter. Trigger Level and Target Level coincide. Grouting is started if
the maximum settlement of all Observation Points exceeds the Trigger Level and continues until
the maximum settlement of all Observation Points reaches the specied Target Level. Note that
Grouting Strategy A only requires that the settlement (i.e. downwards movement) at all
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

COMPENSATION GROUTING ABOVE SHALLOW TUNNELS

s
Ls

DRh = h
Lh
DRs =

DR =
L

s
(a)

Ls

453

Lh

(b)

L/ 2

L/ 2

Figure 7. Deection ratio: (a) deection ratio as dened by Burland and Wroth [29] and Burland et al. [25];
and (b) deection ratio adopted in strategy C.

Observation Points is less than some specied value. Implementation of this strategy is relatively
straightforward and the algorithm is stable. However, note that the strategy could lead to
unacceptably large values of soil heave being generated.
Strategy B. Strategy B is similar to strategy A although it is designed to control soil heave as
well as soil settlement. Grouting is started if the maximum settlement exceeds the Trigger Level
and continues until either the maximum settlement reaches the lower Target Level (which
coincides with the Trigger Level) or the minimum settlement (equal to the maximum heave)
reaches the upper Target Level. This avoids the possibility of grouting inducing substantial
dierential settlements into the structure by generating soil heave.
Strategy C. Strategies A and B use settlement and heave as the SCP, which seems to be
consistent with most current reported site practice. However, discrete measurements of
settlement alone may not always be the best measure of the potential building damage. Strategy
C uses deection ratio as the Settlement Control Parameter; this parameter may provide a more
direct indication of potential building damage.
To compute appropriate values of deection ratio, Observation Points must be grouped along
lines on the ground surface corresponding to the building facades. An arrangement for a simple
four-facade building outline is shown in Figure 8. Observation Points, in groups of three,
provide four deection ratio parameters (DR1 and DR3 for the front and back facades, DR2
and DR4 for the two side facades). The conventional procedure for assessing the likely damage,
outlined for example by Mair et al. [28], considers the part of the building in the sagging region
of the settlement trough separately from the part in the hogging region and is based on the
denition of a sagging DRs and hogging DRh deection ratio (Figure 7(a)). This approach
cannot be used directly in this nite element model because it would involve determining the
point of inection of the settlement curve and adjusting the position of Observation Points
accordingly. Alternatively, a larger number of Observation Points could be used, but this is not
adopted in this paper for the sake of simplicity.
In this model the deection ratio DR between three xed Observation Points is determined, as
shown in Figure 7(b). This deection ratio depends on the position of the Observation Points
and might not necessarily coincide with the denition used by Burland and Wroth [29] and Mair
et al. [28]. However, this is not regarded as a disadvantage if the Observation Points are sensibly
placed.
An additional parameter, termed AVG, is introduced for Strategy C to take account of the
diering eect an increment of grout injection can have on the ground movement beneath
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

454

C. WISSER, C. E. AUGARDE AND H. J. BURD

DR 3

back facade
building

DR 4

DR 2

front facade
Observation point

DR 1

Figure 8. Example for Strategy C (plan view of building outline with 6 observations points and 4
deections ratios DR as Settlement Control Parameter).

dierent facades. AVG is the mean of the absolute values of the observed deection ratios, i.e.
PNDFL
jDRi j
4
AVG i1
NDFL
with NDFL the number of the deection ratios dened by the user
(NDFL would be four for the example shown in Figure 8).
DRi
deection ratio i:
This measure is included to allow for cases where grout injection causes one deection ratio to
increase while the remaining ones are reduced by further grout injection. To illustrate the
importance of AVG, consider the simple building of Figure 8 where further grout injection
would increase one deection ratio above a Target Level. This could be the point to terminate
grout injection. Alternatively one could continue with the injection, as long as the damage in the
other facades was reduced. In this case further damage in one location is accepted, provided the
overall damage is reduced; this is the approach taken for the research presented here. An
additional Target Level is therefore the point where AVG begins to rise, indicating a net
increase in damage for the structure as a whole.
Given the Control Parameters DRi and AVG the control algorithm is as follows, where Nobs
refers to the number of deection ratio measurements:
TA TR
grout off
for each grouting increment do
for i 1 to Nobs do
if DRi > TL then
grout on
end if
end for
if AVG is increasing then
grout off
end if
if grout on then apply grout pressure this step
end for
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

COMPENSATION GROUTING ABOVE SHALLOW TUNNELS

455

4. MODELLING CG FOR TUNNELLING


To illustrate Grouting Strategies A, B and C, several analyses have been carried out using the
mesh shown in Figure 9. The analyses involved the installation of a straight and horizontal
tunnel of length 60 m; diameter 5 m and axis at a depth of 10 m: Due to symmetry, only one half
of the problem was modelled. Overconsolidated clay was modelled using the soil model
described in Section 3.1. The soil parameters employed are given in Table I. Prior to excavation,
initial stresses due to the unit weight of soil and K0 1:25 were generated. A list of the analyses
is given in Table II.

10 m

50 m

z
y
60 m

60 m

boundary conditions: base fixed


vertical faces fixed in direction normal to the face

Figure 9. Mesh for example analyses.

Table II. Example analyses.

Analysis
E1
A1
A2
A3
B1
C1

Grouting
strategy

Grout element shear and normal


stiness kGE
kN=m3

None
A
A
A
B
C

No grout elements inserted


1  105
1  106
1  107
1  106
1  106

Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

456

C. WISSER, C. E. AUGARDE AND H. J. BURD

Observation Points were placed on the ground surface at the possible location of a surface
structure (Figure 10), although note that a structure was not actually modelled in these example
analyses. Further analyses in which buildings are included, are presented by Wisser [11].
Grouting panels were located half way between the tunnel crown and the overlying structure. In
practice, a grouting zone above a tunnel is usually dened by lines extending at 458 from the
tunnel invert [30]. To be consistent with this practice, three grouting panels of dimensions
8:75 m  5 m (denoted 1, 2, 3 in Figure 11) were inserted in the meshes used in these analyses.
Three grouting panels were selected on the initial assumption that this was the minimum
number of independent zones of grouting to provide reasonable control of ground surface
movements. Tunnel excavation and lining installation was carried out in 6 stages, with the

Direction of tunnel construction

CL
5.0 m

5.0 m

rear facade
6

7
5.0 m
assumed building footprint
4
5.0 m

1
3

front facade
Tunnel

Figure 10. Position of the Observation Points (plan view on the ground surface).

12.5 m

12.5 m

5.0 m 5.0 m

12.5 m

12.5 m

direction of tunnelling

excavation stages
grouting panels

8.75 m
building footprint

5.0 m 5.0 m 5.0 m

Figure 11. Plan view of grouting layout.


Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

COMPENSATION GROUTING ABOVE SHALLOW TUNNELS

457

length of one excavation stage being between 5 and 12:5 m: While it is accepted that these
increments are longer than would be adopted in practice, relatively large increments were
adopted in order to keep the run times to acceptable values.
After each increment of excavation, the tunnel liner was shrunk to simulate a ground loss of
2% (see Section 3.1). The model simulated ground loss principally by allowing radial movement
of the soil around the tunnel liner. Face loss is not explicitly modelled. This seems reasonable on
the basis that ground movements around tunnels are conventionally regarded as not being
sensitive to whether the ground loss is associated with the tunnel face, or closure of the soil onto
the lining, or a combination of both. The dierent stages of each analysis are specied in
Table III.
Figure 12 shows surface contours of settlement for analysis E1 at selected stages of
excavation. It is noticeable that ahead of the tunnel face a small amount of heave is created.
Transverse and longitudinal plots of settlement for this analysis are shown in Figure 13
demonstrating that the magnitude of the heave ahead of the tunnel face is small. The surface
settlement trough is, however, wider and shallower than predicted by the semi-empirical
approach [28], as commonly found for numerical analysis of this problem [31].
4.1. Grouting Strategy A
Strategy A is used in analyses A1A3. The limit for the maximum settlement is set to 0 mm with
a tolerance of 1mm: The three analyses in this series are used to investigate the inuence of
initial grout stiness on the nature of results in an analysis with grouting, and dier only in the
initial and nal stiness assigned to the grout elements. Inserting grout elements into a mesh
changes its exibility. The magnitude of this eect depends on the stiness (normal and shear)
assigned to the grout elements. Before CG can be modelled realistically, a suitable stiness for
the grout elements has to be selected. If the stiness is too high, numerical problems caused by
ill-conditioning of the stiness matrix may occur; if it is too low the soil continuum above the
tunnel axis and ahead of the tunnel face may not be modelled correctly.
Figure 14 shows the vertical movements along the surface above the tunnel axis after each
injection stage for analyses A1A3. It is noticeable that signicant heaves are induced in all
analyses; this is a consequence of the Grouting Strategy, which places no control on heave

Table III. Stages of the analyses (see Figure 11 for layout).


Analysis stage
1
2
3n
4
5n
6
7n
8
9n
10
n

Task performed
Excavation to y 12:5 m and shrinking of liner
Excavation to y 25:0 m and shrinking of liner
Grouting of panel 1
Excavation to y 30:0 m and shrinking of liner
Grouting of panels 1 and 2y
Excavation to y 35:0 m and shrinking of liner
Grouting of panels 2 and 3y
Excavation to y 47:5 m
Grouting of panels 3
Excavation to y 60 m and shrinking of liner

Only in analyses A1, A2, A3, B1, C1 where compensation grouting is specied.
Grouting of two panels is dealt with by treating them as a single combined panel.

Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

458

C. WISSER, C. E. AUGARDE AND H. J. BURD

60
12

30
excavated
tunnel

10
0
(a) 0

20

6
4

20

2
10

0
10

30

0
(b) 0

30

60

0
10

20

30

60
14

50

14

50

10
8

30

6
4

20

2
10
0
(c) 0

40
m

40

12
settlements in mm

12

20

6
4
2

10
0
(d) 0

30

60

0
10

20

30

60
14

40

10
8

30

6
4

20

14

50

12
40
m

12

settlements in mm

50

30
20

0
10

10

10
8

30

20

2
10
0
(e) 0

settlements in mm

20

10

0
10

20

30

10
0
(f ) 0

settlements in mm

12
40
m

10
40

14

50
settlements in mm

50

settlements in mm

60

2
10

20

30

Figure 12. Contour plot of surface settlements for E1: (a) after stage 1; (b) after stage 2; (c) after stage 3;
(d) after stage 4; (e) after stage 5; and (f) after stage 6.

magnitude. It should also be noted that the Trigger Level is dened for the Observation Points,
which are located between 25 and 35 m along the tunnel axis. Since the Grouting Strategy is
controlled by the Observation Point movements, settlements are allowed to exceed the Trigger
Level away from these locations. While the surface deformations agree very well between the
three analyses outside the grouting zones, dierences are noticeable above the activated grouting
panels. However, the results of A2 and A3 are suciently close to be assumed that in both cases
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

459

COMPENSATION GROUTING ABOVE SHALLOW TUNNELS

distance from tunnel axis [m]


10
20
30
40
stage 1
stage 2

50

60

50

60

distance along tunnel axis [m]


20
30
40
50

60

stage 2

settlements [mm]

stage 4
stage 3

10

stage 5
15

stage 3

stage 4
10

stage 5

15

stage 6

20

stage 6

20

(a)

(b)
50

60

stage 3
stage 4
stage 5
stage 6

20

0
5
10

e5

settlements [mm]

15

10

stage 1

10

stag

stage 2

distance from tunnel axis [m]


20
30
40

stag
e2
stag
e3
stag
e4

10

e1

stag

settlements [mm]

distance from tunnel axis [m]


20
30
40

10

stage 1

settlements [mm]

stage 6

15

(c)

(d)

Figure 13. Surface deformations for E1: (a) transverse plot at y 25:0 m; (b) transverse plot at y
30:0 m; (c) transverse plot at y 35:0 m; and (d) along tunnel axis.

the eect of grout element exibility on the results is not signicant. In view of this a grout
element stiness of kGE 1  106 kN=m3 (corresponding to run A2) is adopted for later
calculations.
It is useful to assess the additional exibility that grout elements introduce, by comparing
their stiness with the stiness of the surrounding soil. Under purely compressive stress, a grout
element with a normal stiness of kGE experiences the same strain magnitude as a layer of
continuum elements with thickness d and a Youngs modulus E where
d

E
kGE

d can be therefore be thought of as a thickness of soil equivalent to the presence of the grout
elements. The equivalent thickness for analysis A2 is of 0:12 m; using the Youngs modulus of
the continuum elements at the grout panel depth. This seems suciently small to minimise the
additional exibility induced by the grout elements and was therefore used for all subsequent
grouting analyses.
Figure 15 shows the surface settlements for analysis A2 along the tunnel axis and in the
transverse direction at y 25:0; 30.0 and 35:0 m (y is measured along the tunnel axis starting
from the beginning of the tunnel) just before and after the rst injection (analysis stages 2
and 3).The Observation Points are marked on the plots. The inuence of grout injection
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

460

C. WISSER, C. E. AUGARDE AND H. J. BURD

distance along tunnel axis [m]


10

20

40

distance along tunnel axis [m]


50

60
A1
A2
A3

30
settlements [mm]

30

20

10

distance along tunnel axis [m]

settlements [mm]

200

150
100

50

distance along tunnel axis [m]


50

60
A1
A2
A3

600
settlements [mm]

250

40

60
A1
A2
A3

(b) 20

30

50

20

(a) 20

20

40

40

10

30

60

10

20
1

10
0

80
settlements [mm]

40

10

20

30

40
7

400

50

60
A1
A2
A3

200
1
0

(c)

50

(d) 200

Figure 14. Plot of surface settlements along tunnel axis for dierent interface stiness (Grouting Strategy
A): (a) after rst injection (stage 3); (b) after second injection (stage 5); (c) after third injection (stage 7);
and (d) after fourth injection (stage 9).

does not extend far beyond the boundaries of a grouting panel and the heave created decreases
very rapidly towards the panel edge. While the injections cause large heave movements of
Observation Points 1, 2, 6 and 7, which are close to the tunnel axis and above grouting panels,
the eect on Observation Points 3, 4 and 5 (which are located outside the grouting panels, see
Figures 10 and 11) is in comparison very small. Observation Point 3 is always the last of all
Observation Points to reach the specied Target Level, resulting in considerable heave at other
locations closer to the tunnel axis. The maximum heave increases steadily after each injection
stage. As one panel is grouted after the other, Observation Point 3 moves further away from the
currently activated panel. With increasing distance, the eect of a grout injection on an
Observation Point decreases and hence more injection steps are necessary for point 3 to meet
the specied target. This in turn causes considerable heave of the Observation Points closer
to the activated grouting panel. Also of note is the lack of disturbance behind the tunnelling
face from later injections. The contour plots of the surface deformation after each injection
stage, shown in Figure 16, also illustrate some of the observations described above. In particular
the large amount of heave created and the small zone of inuence of the grout injections
are visible.
These results show that poorly devised Grouting Strategies which are based on minimizing
settlement alone, such as Strategy A, can result in large movements which could cause
signicant structural damage. In the case presented here, a considerable amount of unwanted
heave is created and CG would induce more damage than it aims to prevent, if a surface
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

461

COMPENSATION GROUTING ABOVE SHALLOW TUNNELS

distance along tunnel axis [m]


10

20

40

distance from tunnel axis [m]


50

60

Stage2
Stage3
Target
Tolerance

30
settlements [mm]

30

20
10
7

40
settlements [mm]

40

30

10

(a) 20

(b) 10

40

Stage2
Stage3
Target
Tolerance

2
1

tolerance interval
0
4

60

distance from tunnel axis [m]


50

60

1.5

10

20

0.5

30

40

50

60

Stage2
Stage3
Target
Tolerance

1
settlements [mm]

settlements [mm]

30

50
Stage2
Stage3
Target
Tolerance

distance from tunnel axis [m]


20

40

10
0

10

30

20

10

20

7
6

0
5
0.5

(c) 2

(d) 1.5

Figure 15. Surface settlements for A2 after stage 2 (before grouting) and stage 3 (after grouting of panel 1):
(a) plot along tunnel axis; (b) transverse plot at y 25:0 m; (c) transverse plot at y 30:0 m; and
(d) transverse plot at y 35:0 m:

structure were present. The problem is exacerbated by the fact that the spatial arrangement of
grouting panels, involving just three independent regions, is too coarse to achieve satisfactory
control. A more detailed scheme, involving a greater number of independent grouting locations
is described by Wisser [11]. Further discussion of this scheme, however, is beyond the scope of
the current paper.
4.2. Grouting Strategy B
The results for strategy A above clearly demonstrate the need to include heave as well as
settlement control in a Grouting Strategy. Such an approach is contained in Grouting Strategy
B and is now demonstrated using the results of analysis B1. The results of analyses A1A3
indicate that Observation Points away from a grouting panel are little aected by any grouting
in that panel. For this reason only Observation Points above the currently activated panel were
used in the Grouting Strategy algorithm for analysis B1. These are detailed in Table IV.
Contour plots of the vertical surface movements after each injection stage (B1 analysis stages
3, 5, 7, 9) are given in Figure 17. As seen in previous analyses, the grouting eect is localised
above the panels.
Longitudinal and transverse plots of the surface vertical movements for analysis B1 are given
in Figures 1821. Each plot shows the deformations before and after each injection. Since the
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

462

C. WISSER, C. E. AUGARDE AND H. J. BURD

60

60
10

50

10
50

40

10

40

30

20

activated
grouting panel

10

excavated
tunnel

20

20
m

5
30

30

15

50

25

10

40

20

10
60

30
0

10

(a)

20

30

10

(b)

60

20

30

60
0

50

50

40

50

40

50
100

30

100

20

150

20

10

200

10

150
200

30

250
300
350
400
0
(c)

10

20
m

30
(d)

10

20

30

Figure 16. Contour plot of surface settlements (in mm) for A2 after each injection stage: (a) stage 3;
(b) stage 5; (c) stage 7; and (d) stage 9.

Observation Points used by the Grouting Strategy change during the analysis, as dierent panels
are activated, the Target and Trigger Levels are only marked for the Observation Points which
are currently active. Along the tunnel axis (Figures 18(a), 19(a), 20(a), 21(a)) Observation Points
react strongly to the injection of grout and grouting stops when these points hit the heave Target
Level. On the transverse plots it is noticeable that the termination of grouting occurs before all
the Observation Points have been lifted above the maximum settlement level, due to the
restriction on further heave of points over the tunnel axis. As for previous analyses, grout
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

463

COMPENSATION GROUTING ABOVE SHALLOW TUNNELS

Table IV. Observation Points used for dierent grouting panels for analysis B1.
Activated
grouting panel

Analysis
stage

Observation
Points used

3
5
7
9

1, 2, 3
1, 2, 3, 4
4, 5, 6, 7
5, 6, 7

1
1 and 2
2 and 3
3
60

60
10

50

50

10

40

30

20

10

10

5
40

30

activated
grouting panel

20

excavated
tunnel

5
10

10

(a)

15

10

20

30

(b)

60

10

20

30

60
10

50

10

50
5

40

40
0
m

0
5

30

10

20

30

20

10

15
10

0
(c)

10

20

10

20
m

30
(d)

15

10

20

30

Figure 17. Contour plot of surface settlements (in mm) for B1 after each injection stage: (a) stage 3;
(b) stage 5; (c) stage 7; and (d) stage 9.
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

464

C. WISSER, C. E. AUGARDE AND H. J. BURD

distance along tunnel axis [m]


10

20

30

40

distance from tunnel axis [m]


50

60

Stage2
Stage3
Target
Tolerance

10

20

15
settlements [mm]

settlements [mm]

20

0
10

10

10

20

30

40

50

60

Stage2
Stage3
Target
Tolerance

1
2

5
0

(a)

20

(b)

10

distance from tunnel axis [m]

(c)

10

20

30

40

1
0
1
2

distance from tunnel axis [m]


50

60

0.5

Stage2
Stage3

10

20

30

40

50

60

Stage2
Stage3
settlements [mm]

settlements [mm]

(d)

0.5

Figure 18. Surface settlements for B1 after stage 2 (before grouting) and stage 3 (after grouting of panel 1):
(a) plot along tunnel axis; (b) transverse plot at y 25:0 m; (c) transverse plot at y 30:0 m; and
(d) transverse plot at y 35:0 m:

injection in the panels used here has very little eect on Observation Points 3 and 4. Another
observation of the model behaviour which seems to agree with site experience is that once
grouting has been completed in a panel, and the tunnel is advanced, there is negligible change to
surface movements behind the face (see Figures 20(b), 21(b), 21(c)).
Strategy B avoids the generation of large heaves and therefore represents an improvement on
Strategy A. It cannot, however, ensure that settlements are kept below Target Levels because the
constraints on settlements and heaves conict.
Maximum settlement and heave cannot be satised by as few as three grout panels and one
independent grout pressure. In practice, of course, a large number of grout injection points
would be used, each with independent values of grout pressure. The use of an increased number
of independent grout injection pressures in the model would clearly improve its ability to control
the surface settlements. This would increase considerably the complexity of the control
algorithm used to determine the optimum grout pressures, however, and is beyond the scope of
the current paper.
4.3. Grouting Strategy C
The results of analyses A2 and B1 have shown that control using absolute deection
measurement may not provide adequate control of building damage, which is the objective of
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

465

COMPENSATION GROUTING ABOVE SHALLOW TUNNELS

distance along tunnel axis [m]


10

20

30

40

60

Stage4
Stage5
Target
Tolerance

1
settlements [mm]

distance from tunnel axis [m]


50

10

20
settlements [mm]

20

0
10

10

20

30

40

50

60

Stage4
Stage5
Target
Tolerance

15 1
10
2
5
0
3

(a) 20

(b) 5
distance from tunnel axis [m]
20

30

40

distance from tunnel axis [m]


50
Stage4
Stage5
Target
Tolerance

15
10
5
0

(c)

10

60

0.5

10

20

30

40

50

60

Stage4
Stage5
settlements [mm]

settlements [mm]

20

0
0.5
1

(d) 1.5

Figure 19. Surface settlements for B1 after stage 4 (before grouting) and stage 5 (after grouting of panel 1
and 2): (a) plot along tunnel axis; (b) transverse plot at y 25:0 m; (c) transverse plot at y 30:0 m; and
(d) transverse plot at y 35:0 m:

any CG operation. Better control over building damage may therefore be possible with
Grouting Strategy C which uses deection ratio instead of absolute movement to control the
grouting. Analysis C1 described below was undertaken to investigate this strategy.
The deection ratios of the three facades making up the notional building were the SCP for
Grouting Strategy C in analysis C1. Using symmetry (see Figure 22), the deection ratios were
calculated as follows:
s1  s3
DR1
w
s4  12 s3 s5
d
s7  s5
DR3
w

DR2

where DR1 ; DR2 ; DR3 are the deection ratios of the notional front, side and rear facades
respectively (the front facade being the one passed rst by the tunnel). s1 to s7 are the settlements
of Observation Points 17 (downwards movements correspond to a positive settlement) and
w and d are the width 20 m and depth 10 m of the building (see Figure 10). With this
denition a positive deection ratio corresponds to a sagging mode, and a negative value
corresponds to a hogging mode.
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

466

C. WISSER, C. E. AUGARDE AND H. J. BURD

distance along tunnel axis [m]


10

20

30

40

distance from tunnel axis [m]


50

60

Stage6
Stage7
Target
Tolerance

20
1
10
0

15
settlements [mm]

settlements [mm]

30

10

(b)

10

(a) 20

10

20

30

40

Stage6
Stage7
Target
Tolerance

15
10
5
0
4

(c) 5

40

50

60

5
0

distance from tunnel axis [m]


50

60

20

10

10
5
0

20

30

40

50

60

Stage6
Stage7
Target
Tolerance

15 7
settlements [mm]

settlements [mm]

20

10

30

Stage6
Stage7

distance from tunnel axis [m]


0

20

6
5

(d) 10

Figure 20. Surface settlements for B1 after stage 6 (before grouting) and stage 7 (after grouting of panel
2 and 3): (a) plot along tunnel axis; (b) transverse plot at y 25:0 m; (c) transverse plot at y 30:0 m;
and (d) transverse plot at y 35:0 m:

In analysis E1 the maximum deection ratio, as dened by Equation (6) was approximately
0.04% and occurred in stage 4 between Observation Points 7 and 5 DR3 : This information was
used to set the Target Level for analysis C1, i.e. to limit the maximum deection ratio to one half
of this value (0.02%).
Values of deection ratios DR1 ; DR2 and DR3 during the analysis are plotted in Figure 23.
This gure shows that at the end of stage 2 (i.e. at the start of stage 3) all deection ratios are
positive (indicating sagging deformations). The deection ratio of the front facade DR1
slightly exceeds the Trigger Level of 0.02% and so grout is injected, in three steps, until DR1
falls below 0.02% (see Figure 23(a)).
Due to tunnel excavation in the next stage (stage 4) the deection ratio of the front facade
increases considerably and, at the end of this stage, it reaches a value of 0.038%. Since this
exceeds the Trigger Level of 0.02%, further grout injection is modelled in stage 5 (see Figure
23(b)). This grouting procedure is seen to be eective in reducing DR1 and, the grout injection
terminates when DR1 falls below 0.02%. (Note that in this analysis the nal value of DR1 is
0.014%. Although the algorithm ensures that the nal deection ratio is less than the target
value, the nite value of grout pressure increment, Dp; means that the values of deection ratio
may undershoot the target value). The grouting procedure modelled in Stage 5 also has the eect
of increasing, slightly, the deection ratio in the rear facade DR3 : However, this deection ratio
remains below the Trigger Level of 0.02%.
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

467

COMPENSATION GROUTING ABOVE SHALLOW TUNNELS

distance along tunnel axis [m]


10

20

30

40

60

Stage8
Stage9
Target
Tolerance

1
settlements [mm]

distance from tunnel axis [m]


50

10
0
10

(a) 20

15
settlements [mm]

20

10

(b)

10

20

30

40

60

50

60

5
0

20

Stage8
Stage9

15

40

distance from tunnel axis [m]


50

settlements [mm]

settlements [mm]

20

10

30

Stage8
Stage9

distance from tunnel axis [m]


0

20

10
5
0

10

20

30

40

50

60

Stage8
Stage9
Target
Tolerance

15 7
10
6

5
0

(c) 5

(d)

5
5

Figure 21. Surface settlements for B1 after stage 8 (before grouting) and stage 9 (after grouting of panel 3):
(a) plot along tunnel axis; (b) transverse plot at y 25:0 m; (c) transverse plot at y 30:0 m; and
(d) transverse plot at y 35:0 m:

symmetry axis
L /2

observation point 3

observation point 1

Figure 22. Average deection ratio DR1 of the front facade using symmetry.

At the end of stage 8 (i.e. at the start of stage 9) it is clear from Figure 23(c) that the deection
ratio in the rear facade exceeds the exceeds the Trigger Level. This results in a further grout
injection in stage 9 that is seen to reduce the deection ratio to below the Target Level of 0.02%.
The developments of deection ratios DR1 ; DR2 and DR3 during all analyses discussed in
this paper are plotted, with respect to excavated tunnel length, in Figure 24. It is noticeable that
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

468

C. WISSER, C. E. AUGARDE AND H. J. BURD

DR1
DR
2
DR3

deflection ratio (%)

0.02
0.01

AVG
limit

0
0.01
0.02
1

1.5

2
step

(a)

2.5

deflection ratio (%)

0.04

0.02

DR
1
DR2
DR3

0.01

AVG
limit

0.03

0
0.01
0.02
2

10

(b)

12

14

16

18

0.04
deflection ratio (%)

20

step

0.02

DR1
DR2
DR

0.01

AVG
limit

0.03

0
0.01
0.02

(c)

10

15

20

step

Figure 23. Development of deection ratio for analysis C1 in grouting stage 3, 5 and 9 (in stage 7 no
grouting because all Target Levels are met). Note: AVG jDR1 j jDR2 j jDR3 j=3: (a) stage 3;
(b) stage 5; and (c) stage 9.

Strategy A (analysis A2) results in large deection ratios which would induce more damage to a
building rather than prevent it. Limiting the amount of heave with strategy B (analysis B1) leads
to smaller deection ratios. However, since the deection ratios at the end of analysis B1 exceed
those of analysis E1 (an analysis with no compensation grouting), grouting also in this case may
increase rather than reduce any damage to a surface structure. More promising is strategy C
(analysis C1) which provides smaller values of deection ratio, in particular for DR1 ; than
analysis E1 where no grouting is carried out. This suggests that strategies based on deection
ratio, rather than absolute deection, may be more successful in protecting buildings from
tunnelling induced damage.
4.4. Analyses involving a masonry building on the soil surface
Further calculations in which a masonry building is explicitly included in the model are
described by Wisser [11]. In these calculations an attempt was made to use Grouting Strategy C
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

deflection ratio (%)

COMPENSATION GROUTING ABOVE SHALLOW TUNNELS

0.10
0.05
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40

E1
A2
B1
C1

tunnel length excavated (m)

(a)

deflection ratio (%)

469

0.010
0.005
0.000
0.005
0.010
0.015
0.020
0.025
0.030
0.035

(b)

E1
A2
B1
C1

tunnel length excavated (m)

deflection ratio (%)

0.5
0.0
0.5

E1
A2
B1
C1

1.0
1.5
2.0
2.5

(c)

tunnel length excavated (m)

Figure 24. Development of deection ratio for analysis E1, A2, B1, C1 (in analyses A2,
B1, C1 grouting at an excavated tunnel length of 25.0, 30.0, 35.0 and 47:5 m): (a) DR1
(front); (b) DR2 (side); and (c) DR3 (back).

to minimize the predicted damage to the building. Although, in this modelling exercise the
grouting operations were able to reduce the values of deection ratio, the control of
deformations in a building was found to be a substantially more dicult task than that of
controlling deformations at the ground surface when a building is absent. Activation of a
particular grout panel could be used to reduce the deection ratio in one of the facades, but this
often had the eect of increasing it in other facades. This behaviour seems to be caused by the
building providing a sti coupling between the Observation Points. It is thought that, in
principle, a numerical model of compensation grouting procedures for the case where a building
is placed on the ground surface could be developed, but this would require rather more than the
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

470

C. WISSER, C. E. AUGARDE AND H. J. BURD

three independent grout panels and a single independent grout pressure as adopted in this
model. An advanced model of this sort has not yet been attempted.

5. CONCLUSIONS
A three-dimensional numerical model of compensation grouting above a tunnel in overconsolidated clay has been presented. Various dierent approaches to controlling the injection
of grout are possible in compensation grouting depending on Control Parameters that are
adopted. Since building damage generally needs to be minimized it appears sensible to adopt a
control strategy based on a parameter that is closely related to accepted measures of damage.
This numerical model has been shown to replicate general behaviour expected from the eld.
The model is based on three independent panels to model the grout injection process. This
approach provides a model that is capable of controlling ground surface movements for the case
where a surface building is absent. Equivalent analyses (not reported in detail in this paper)
suggest that a more sophisticated approach involving an increased number of independent grout
panels and a more complex control algorithm, is needed for cases where a building exists at the
ground surface.

ACKNOWLEDGEMENTS

The rst author was supported by an EPSRC studentship during the period of this research. Some of the
calculations presented were performed at the Oxford Supercomputing Centre.
REFERENCES
1. Harris DI, Mair RJ, Love JP, Taylor RN, Henderson TO. Observations of ground and structure movements for
compensation grouting during tunnel construction at Waterloo Station. Ge!otechnique 1994; 44(4):691713.
2. Soga K, Bolton MD, Au SKA, Komiya K, Hamelin JP, Van Cotthem A, Buchet G, Michel JP. Development of
compensation grouting modelling and control system. In Proceedings of the International Symposium on Geotechnical
Aspects of Underground Construction in Soft Ground, Kusakabe O, Fujita K, Miyazaki Y (eds), Tokyo, Japan, 1921
July, 1999. Balkema: Rotterdam, 2000; 425436.
3. Nicholson DP, Gammage C, Chapman T. The use of nite element methods to model compensation grouting.
Grouting in the Ground, November 1994, The Institution of Civil Engineers. Thomas Telford: London, 1994;
297312.
4. Mayer P-M, Kudella P. Prognose von Bodenverformungen als Bestandteil der Beobachtungsmethode. Sonderdruck
Zur 4. FEM/CAD, Tagung, Darmstadt. RIB Bausoftware GmbH, Marketing RIBTEC, Vaihingerstr. 151, 70567,
Stuttgart, Germany, 1996; 1319.
5. Kovacevic N, Edmonds HE, Mair RJ, Higgings KG, Potts DM. Numerical modelling of the NATM and
compensation grouting trials at redcross way. In Geotechnical Aspects of Underground Construction in Soft Ground,
Mair RJ, Taylor RN (eds), 1517 April, 1996, City University, London. Balkema: Rotterdam, 1996; 553559.
6. Standing JR, Farina M, Potts DM. The prediction of tunnelling induced building settlements}a case study. In
Proceedings of the World Tunnel Congress98 on Tunnels and Tunnels and Metropolises, Negro Jr A, Ferreira AA
(eds), Sao Paulo, Brazil, 2530 April, 1998. Balkema: Rotterdam, 1998; 10531058.
7. Falk E, Schweiger HF. Shallow tunnelling in urban environment}dierent ways of controlling settlements. Felsbau
1998; 16(4):215223.
8. Schweiger HF, Falk E. Reduction of settlements by compensation grouting}numerical studies and experience from
Lisbon underground. Proceedings of the World Tunnel Congress98 on Tunnels and Metropolises, Sao Paulo, Brazil,
2530 April, 1998. Balkema: Rotterdam, 1998; 10471052.
9. Buchet G, Van Cotthem A. 3D steady state numerical modeling of tunneling and compensation grouting. FLAC
and Numerical Modeling in Geomechanics, Proceedings of the International FLAC Symposium, Minneapolis,
September 1999. Balkema: Rotterdam, 1999; 255261.
Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

COMPENSATION GROUTING ABOVE SHALLOW TUNNELS

471

10. Addenbrooke TI, Ong JCW, Potts DM. Finite-element analysis of a compensation grouting eld trial in soft clay.
Proceedings of the Institution of Civil Engineers, Geotechnical Engineering 2002; 115(1):4758.
11. Wisser C. Numerical modelling of tunnel installation and compensation grouting. D.Phil. Thesis, Oxford University,
2002.
12. Nicolini E, Nova R. Modelling of a tunnel excavation in a non-cohesive soil improved with cement injections.
Computers and Geotechnics 2000; 27:249272.
13. Burd HJ, Houlsby GT, Augarde CE, Liu G. Modelling tunnelling-induced settlement of masonry buildings.
Proceedings of the Institution of Civil Engineers, Geotechnical Engineering 2000; 143:1729.
14. Houlsby GT. A model for the variable stiness of undrained clay. Proceedings of the International Symposium on
Pre-Failure Deformations of Soil, Torino, 2629 September. Balkema: Rotterdam, 1999; 443450.
15. St John HD, Potts DM, Jardine RJ, Higgins KG. Prediction and performance of ground response due to
construction of a deep, basement at 60 Victoria Embankment. In Predictive Soil Mechanics, Proceedings of the
Wroth Memorial Symposium, Houlsby GT, Schoeld AN (eds), held at St Catherines College Oxford, 2729 July,
1992. Thomas Telford: London, 1993; 581608.
16. Kim S. Model testing and analysis of interaction between tunnels in clay. D.Phil. Thesis, University of Oxford, 1996.
17. Ng CWW, Lings ML, Simpson B, Nash DFT. An approximate analysis of the three-dimensional eects of
diaphragm wall installation. Ge!otechnique 1995; 45(3):497507.
18. Liu G. Numerical modelling of settlement damage to masonry buildings caused by tunnelling. D.Phil. Thesis,
University of Oxford, 1997.
19. Augarde CE, Burd HJ. Three-dimensional nite element analysis of lined tunnels. International Journal for
Numerical and Analytical Methods in Geomechanics 2001; 25:243262.
20. Brown PT, Booker JR. Finite element analysis of excavation. Computers and Geotechnics 1985; 1:207220.
21. Edmonds H. Control and restraint. Ground Engineering 1997; March:3841.
22. Hubbert MK, Willis DG. Mechanics of hydraulic fracturing. Transactions of the American Society of Mining,
Metallurgy and Petroleum Engineers 1957; 210:153168.
23. Ngo-Tran CL. The analysis of oshore foundations subjected to combined loading. D.Phil. Thesis, University of
Oxford, 1997.
24. Houlsby GT, Liu G, Augarde CE. A tying scheme for imposing displacement constraints in nite element analysis.
Communications in Numerical Methods in Engineering 2000; 16:721732.
25. Burland JB, Broms BB, de Mello VFB. Behaviour of foundation and structures. Proceedings of the 9th International
Conference on Soil Mechanics and Foundation Engineering, Tokyo, 1977; 495546.
26. Boscardin MD, Cording EJ. Building response to excavation-induced settlement. Journal of Geotechnical
Engineering (ASCE) 1989; 115(1):121.
27. Burland JB, Standing JR, Jardine FM (eds). Building Response to Tunnelling, Case Studies from Construction of the
Jubilee Line Extension, vol. 1. Thomas Telford: London, 2001.
28. Mair RJ, Taylor RN, Burland JB. Prediction of ground movements and assessment of risk of building damage due
to bored tunnelling. In Geotechnical Aspects of Underground Construction in Soft Ground, Mair RJ, Taylor RN (eds).
Balkema: Rotterdam, 1996; 713718.
29. Burland JB, Wroth CP. Settlement of buildings and associated damage. Settlement of Structures, Conference,
organised by the British Geotechnical Society at the Lady Mitchell Hall, Cambridge. British Geotechnical society,
Pentech: London, 1974; 611654.
30. Haimoni AM. Tunnelling with compensation grouting, Jubilee line extension. Going Underground 97, National
Conference, Birmingham, paper four, 23 October, 1997, The Institution of Civil Engineers, 1997.
31. Mair RJ, Taylor RN. Theme lecture: bored tunnelling in the urban environment. Proceedings of the Fourteenth
International Conference on Soil Mechanics and Foundation Engineering, Hamburg, 612 September, 1997. Balkema:
Rotterdam, 1997.

Copyright # 2005 John Wiley & Sons, Ltd.

Int. J. Numer. Anal. Meth. Geomech. 2005; 29:443471

You might also like